SLaTE 2015 - Workshop on Speech and Language Technology in Education

Leipzig, Germany
September 4-5, 2015

Towards a Conversational Expert System for Rhetorical and Vocal Quality Assessment in Call Center Talks

Mathias Walther (1), Baldur Neuber (2), Oliver Jokisch (3), Taïeb Mellouli (1)

(1) Department of Business Information Systems and Operations Research, Martin Luther University Halle-Wittenberg, Germany
(2) Department of Speech and Phonetics, Martin Luther University Halle-Wittenberg, Germany
(3) Institute of Communications Engineering, Leipzig University of Telecommunication, Germany

This article presents the concept and development steps towards a conversational expert system for rhetorical and vocal quality assessment in call center talks. At first the state of the art in quality assessment is discussed. The influencing rhetorical and vocal factors are introduced. In our novel approach, the recognition of vocal factors is modeled by competing classification systems and combined into a multi-classifier system which is based on decision trees. Finally we propose an expert system which incorporates the generated decision rules. The system accuracy can be improved by user-adapted rule sets. Furthermore solutions to the problem of inconsistent rules are discussed.

Full Paper

Bibliographic reference.  Walther, Mathias / Neuber, Baldur / Jokisch, Oliver / Mellouli, Taïeb (2015): "Towards a conversational expert system for rhetorical and vocal quality assessment in call center talks", In SLaTE-2015, 29-34.